Geoffrey Hinton Warns of 10% AI Extinction Risk, Narrowing Timeline to Under a Decade

Translation Summary

Okay, so the task is to translate this Chinese news article about Geoffrey Hinton's remarks on AI existential risk. The goal is a professional translation, maintaining the original paragraph structure and URLs, with no personal commentary.

First, I need to analyze each paragraph.

• Paragraph 1: The article starts with Hinton's view: it's hard to estimate the probability of AI causing human extinction, but it's not zero. Simple enough, just need to make sure the nuance of the Chinese phrasing is captured accurately in the English translation.

• Paragraph 2: This is the core quote from the Newsnight interview. The translation needs to be precise. It seems a 10% chance isn't considered ridiculous. I'll need to check the exact wording of the original BBC quote to ensure accuracy.

• Paragraph 3: This is where it gets interesting. I need to make sure the researcher's name from Anthropic is accurate. It's Jacob Coxon. The article seems to confirm Hinton's agreement with Coxon's assessment: a "more than 10% chance of killing everyone" within 10 years.

• Paragraph 4: This just seems to be a reaction from the interviewer to Hinton and Coxon.

• Paragraph 5: It is important to translate this as "may very soon be smarter than us".

• Paragraph 6: Hinton considers 1% as unlikely. I should ensure that this is the meaning translated.

• Paragraph 7: Hinton's timeline is shortening significantly, going from 30-50 years, to 10-20, now maybe even 10 or less. The original language has to reflect this narrowing.

• Paragraph 8: The reasons for the timeline's shortening are broad: Cyberattacks or AI developing dangerous viruses. The recent advancements in AI models and investment from companies like Anthropic and OpenAI are key.

• Paragraph 9: It mentions Jacob Coxon's resignation from Anthropic prior to an IPO. This is important context, particularly his previous role at OpenAI and his strong statement about these companies "gambling with our lives".

• Paragraph 10: The quotes from Coxon on X are concerning. This confirms the concern from AI professionals.

• Paragraph 11: Hinton presents both potential scenarios: self-destruction or finding a way to mitigate risks. Global regulation is presented as a means to the second solution.

• Paragraph 12: Countries have aligned interests in preventing AI from replacing humanity, which will lead to collaboration. The key question is whether they will act in time.

I'll now translate each paragraph based on these considerations. Then, I will review the translation to ensure accuracy and professionalism.

On September 11, Beijing time, according to Business Insider, Geoffrey Hinton, the "Godfather of AI," stated that while it is difficult to estimate the probability of AI causing the total extinction of humanity, the chance is not zero.

Speaking to BBC's Newsnight on Wednesday, Hinton said: "Ten percent seems like a not crazy estimate to me, but nobody actually knows how to make a sensible estimate."

Newsnight presenter Victoria Derbyshire expressed shock, as Hinton essentially agreed with the assessment of Jacob Coxon, a top AI safety researcher at Anthropic. The researcher had stated that within the next 10 years, AI has "a more than 10% chance of killing everyone."

"Wow," Derbyshire said. "My goodness."

Hinton noted that it is difficult to estimate an exact probability because human society has never encountered a technology that "may very soon be smarter than us."

"It would be silly to say it was 1%," he said. "Nobody knows how to estimate it."

The Nobel laureate said what is certain is that the window for such a catastrophe to potentially happen has narrowed. "I used to think it was 30 to 50 years, then I revised it down to 10 to 20 years," he said. "Now, I think it could be 10 years, or even less. And other people think it's much faster than that."

As for how this scenario could unfold, Hinton cited risks ranging from cyberattacks to AI developing lethal viruses. Although Hinton did not specify why the timeline had shrunk so drastically, leading AI companies—including Anthropic and OpenAI—have invested billions of dollars into scaling up models, with their latest releases demonstrating marked improvements over systems from just months earlier.

Hinton's comments followed the resignation of Anthropic researcher Coxon ahead of the company's anticipated IPO. Coxon, who previously worked at OpenAI, said these AI giants are "gambling with our lives."

"People building AI genuinely believe it could kill us all before the decade is out," Coxon wrote on X. "This is not marketing hype. If anything, many executives and senior researchers water down their public statements to sound reasonable, but in private I hear the same people expressing terror."

Hinton said humanity could head down a path of self-destruction, or find ways to "cope with the dangers of AI" while realizing its potential. He reiterated his call for stronger global regulation, noting that while countries may not agree on everything, they share a common desire to preserve human survival.

"When it comes to preventing AI from replacing humanity, countries' interests are aligned. So they will eventually cooperate. The question is, will they act in time?" he said.